Baltimore Judge declares Fingerprints not reliable.

[RedFive]

9/4/09 update:
http://www.explorebaltimorecounty.com/n ... alty-case/

[Outsider]

It seems to me that there is a lot of bad reasoning in this case.

At today’s motions hearing, McCullough, who argued Rose’s position, said that the Mayfield case is the perfect example of why fingerprint evidence should not be allowed in court cases.

“Following this process does not ... produce reliable results,” McCullough said.

One error does not mean that results, in general, are not reliable. Even if there are many more errors than Mr/Ms McCullough knows about this does not stop fingerprint evidence being of a very high probative value in most cases.

Blake said she knew that fingerprint evidence is not “100 percent” and that “there’s always some sort of error rate and people make mistakes.”

Purcell said the few errors in the vast body of fingerprint evidence are “statistically insignificant,” though he conceded: “It’s not insignificant if you’re Mr. Mayfield.”

I doubt very much if Blake and Purcell have any idea what “statistically significant” means. An extremely small general likelihood of forensic error may not result in an extremely small likelihood of error in the context of a particular case.

“There’s never, ever been anywhere where two people had the same fingerprint,” the prosecutor said.

No two deposits from the same finger produce identical fingerprints. The question is in what circumstances, and to what degree, can a finder of fact rely on the evidence of fingerprint experts.

Of course all we have is the newspaper report of the debate but I am wondering if the adversarial system will ever get beyond simplistic generalisations about the value of evidence. It concerns me that the police and others who have to evaluate evidence are being persuaded to think in black and white. Fingerprint evidence is either 100% reliable or it is useless.

We now know that Shirley McKie was prosecuted on the basis that fingerprints were “reliable” and the fact that a fingerprint had never previously been successfully challenged in court. These arguments don’t impress me much. Commercial aircraft are extremely reliable these days but if I am out walking and I see wreckage that has a fuselage wings and engines you will have a hard job persuading me that this could not be a plane crash because aircraft are reliable.

Identification evidence is generally very reliable so I don’t see why it should be excluded from the total package of evidence against someone. But if no other evidence against the identified person can be found, and there is nothing else to link this person to the crime, maybe forensic error should be considered as a possible explanation for this. Jury members can not be expected to be able to estimate the likelihood of forensic error precisely (which they would need to do if there is no other evidence against the accused) so somebody in the justice system should be worried that this would be a risky prosecution.

[Michele]

The question is in what circumstances, and to what degree, can a finder of fact rely on the evidence of fingerprint experts.

I agree, before we decide how reliable the evidence is on a general level, I think it’s important to figure out when this evidence is reliable. The question becomes, what are the determining factors, or relevant variables, in determining reliability?

It sounds like you are saying that ‘other evidence’ in the case should determine the weight of the forensic evidence? What kind of ‘other evidence’ do you think would support forensic evidence?

[Outsider]

Hi Michele,

It sounds like you are saying that ‘other evidence’ in the case should determine the weight of the forensic evidence? What kind of ‘other evidence’ do you think would support forensic evidence?

I don’t know the answer to this question any more than anyone else. But since we do not know, and probably never will know, the true general likelihood error for fingerprint identification (or DNA identification or any other technique where humans are involved) I would say that great risks are being taken if there is NOTHING else to link the identified person to the crime. Anyone who becomes a suspect independently from the forensic evidence or anyone who is even known to the victim has some prior connection with the crime.

Some time ago “g.” suggested on this forum that proximity to a crime increases the likelihood that someone committed it. If we assume that fingerprint evidence is generally extremely reliable (which I think it is) then some degree of proximity might be enough. Imagine a national database search that hits on someone who lives in the same street or district as the crime. A random error during a database search is much more likely to hit on someone who lives far away and has no connection to the crime (someone in Oregon being identified for a crime in Madrid for example).

However I don’t think proximity is enough in the McKie case. After an ID has been disputed the likelihood of it being erroneous is linked (inversely) to the likelihood that the suspect is lying. There must be a reason for someone to lie and the crime is usually that reason. Proximity to a crime with a disputed fingerprint in the location is a very different matter from proximity to just a disputed fingerprint (no matter what that fingerprint infers).

By the way the above only applies to errors where the effects are random. Errors due to bias or down right badness are going to happen to people who are already believed to be the perpetrator of the crime. There is no way to isolate IDs that are at a high risk of bias after they have occurred because they are indistinguishable from good IDs (which tend to hit on people where there is other evidence that they did the crime, because they DID do the crime).

[Michele]

I see 3 different issues being raised, 1) when is an conclusion at risk? 2) When is a conclusion correct? And 3) what significance the conclusion holds to prosecution? Although these issues aren’t mutually exclusive, they are independent of each other and have different variables.

Issue 1
One determinant of when a conclusion is ‘at risk’ of being incorrect may be when two experts arrive at different conclusions. It’s hard to look at known cases (Mckie or Mayfield) because we don’t know whether or not any examiners initially disagreed with the conclusion, but they didn’t seem to have compelling agreement after the results were released. Just because disagreement didn’t exist in the initial reports doesn’t mean that disagreement didn’t exist.

Another determinant in ‘at risk’ conclusions is when the person being ID’d protests the conclusion (or wasn’t in the proximity). This may determine if the person had the ability to leave a print, but it doesn’t determine if the print in question was left by a certain individual. As you indicated, disputing a conclusion doesn’t mean an error occurred, it’s just one reason that a conclusion might want to be reviewed.

Another determinant in ‘at risk’ conclusions may be when the quality is poor and the quantity of information is poor. I’ve seen a lot of errors but I’ve never seen one when the quality and quantity is high. In jails, a lot of people dispute that they are the right person being arrested on warrants, but this is easy to determine by comparing tenprint records. Once these records are compared, I’ve never seen an inmate continue to dispute the ID. I’d guess that this is because the quality and quantity with tenprint to tenprint ID’s are so high that even the inmate acknowledges the conclusion.

Issue 2
A correct conclusion is dependent on the amount of agreement (and disagreement) between the images (and not dependent on if other evidence exists against a person or on whether or not they were in proximity to the object being touched). The amount of agreement of disagreement may vary depending on the quality of the images. The problems with Mckie and Mayfield cases are about the comparisons themselves, all the other information being discussed (as interesting as it may be) is irrelevant to whether or not a valid ID was arrived at.

Issue 3
LPE’s are not prosecutors. It’s up to the prosecutor to decide when and if additional evidence is needed to press charges in a case. If a person was stabbed with a knife, and then a wet bloody print was found on the knife and identified to a suspect, what other evidence does the prosecution need? They don’t need additional evidence to justify the identification, they just need overwhelming agreement between the patent print and the known print to validate the identification.


The Mayfield case applies to both an ‘at risk’ case and an erroneous ID, so it’s natural to look at all the determinants and think they apply to all the issues, when in fact they don’t. The Mayfield print wasn’t erroneous because he disputed the conclusion, or because the Spanish experts didn’t agree, it was erroneous because it lacked sufficient agreement between the latent print and the known print. The fact that it was disputed didn’t mean it was an error, it just opened it up for review so the error could be found.

[Outsider]

Hi Michele,

One determinant of when a conclusion is ‘at risk’ of being incorrect may be when two experts arrive at different conclusions. It’s hard to look at known cases (Mckie or Mayfield) because we don’t know whether or not any examiners initially disagreed with the conclusion

It is an interesting idea that the degree of disagreement within the fingerprint department could be a predictor of a risky ID.

In the McKie case we know that some SCRO experts (Dunbar, Bruce and Foley but I think there are more) could not find the 16 points required for court identification but could find enough to “eliminate” the print as McKie’s. All (except Foley I think) are still completely convinced it is McKie’s print but they could not testify in court.

Malcolm Graham is independent of the SCRO and was employed by David Asbury’s legal team to look at a number of fingerprints including Y7. He confirmed that Y7 was McKie’s print even though the could not get 16 points. He knew that this ID was used in court yet he did not think it necessary to tell the people who commissioned his work that his confirmation was not to the court standard. In his submission to the current Inquiry Graham said that you could identify someone “with a very high degree of probability” with 6 or 7 points (FI_0089).

These experts seem to be using some sort of non-numeric standard to arrive at their confident conclusions from the images alone. You might remember Malcolm Graham had some forthright things to say at the last inquiry about someone who had a different conclusion.

http://news.bbc.co.uk/1/hi/scotland/gla ... 114456.stm

Perhaps at the end of his Inquiry Sir Anthony Campbell will tell us who he thinks the "charlatans" are.

[Heather Baxter]

Judge Blake ruled to admit fingerprint evidence without holding a Daubert hearing.

Back in 2007, Judge Souder ruled to exclude fingerprint evidence in the state case, MD v. Rose. The case was then moved to Federal Court where Daubert motions were filed by the defense. The Judge ruled today to admit the fingerprint evidence at the Federal trial, without holding a Daubert hearing.

IN THE UNITED STATES DISTRICT COURT
FOR THE DISTRICT OF MARYLAND
UNITED STATES OF AMERICA

v.
BRIAN KEITH ROSE

CRIMINAL NO. CCB-08-0149

ORDER
For reasons to be more fully explained in a memorandum opinion, I have concluded that
no further evidentiary hearing is required for the court to find fingerprint identification testimony
admissible in this case under the standard set by the Supreme Court in Daubert v. Merrell Dow
Pharmaceuticals, Inc., 509 U.S. 579 (1993) and Kumho Tire Co., Ltd. v. Carmichael, 526 U.S.
137 (1999). Accordingly, the hearing set for September 30, 2009 is Cancelled.
SO ORDERED this 8th day of September, 2009.

Catherine C. Blake
United States District Judge
Case 1:08-cr-00149-CCB Document 56 Filed 09/08/2009 Page 1 of 1